CN113782146A

CN113782146A - General medicine recommendation method, device, equipment and medium based on artificial intelligence

Info

Publication number: CN113782146A
Application number: CN202111086398.1A
Authority: CN
Inventors: 徐啸
Original assignee: Ping An Technology Shenzhen Co Ltd
Current assignee: Ping An Technology Shenzhen Co Ltd
Priority date: 2021-09-16
Filing date: 2021-09-16
Publication date: 2021-12-10
Anticipated expiration: 2041-09-16
Also published as: CN113782146B

Abstract

The application discloses a general medication recommendation method, device, equipment and medium based on artificial intelligence, relates to the technical field of artificial intelligence and digital medical treatment, and can solve the technical problem of low accuracy of a medication recommendation model. The method comprises the following steps: determining a sample set constructed by sample pathological data, wherein the sample set comprises a first sample set without noise data after noise detection and a second sample set without noise detection; the method comprises the steps of pre-training a first medicine recommendation model by using a first sample in a first sample set, inputting a second sample in a second sample set after pre-training is finished, obtaining the prediction probability and prediction uncertainty of a first medicine recommendation result, screening a first preset number of second samples in the second sample set according to the first preset number of second samples to carry out noise detection, iteratively training the first medicine recommendation model by using the second sample without noise data after the noise detection and the first sample set to obtain a target medication recommendation model, and obtaining a target recommended medication and sending the target medication recommendation model to a target visiting patient according to the target medication recommendation model.

Description

General medicine recommendation method, device, equipment and medium based on artificial intelligence

Technical Field

The application relates to the technical field of artificial intelligence, in particular to a general medication recommending method, device, equipment and medium based on artificial intelligence.

Background

The general medication recommendation means that reasonable medication is automatically recommended for visiting patients in general scenes according to the injury conditions (diagnosis, inspection, and the like). Can reduce unreasonable medication caused by shallow qualification of doctors, also prevent the phenomenon of excessive prescription, is beneficial to the symptomatic administration of patients, improves the medical quality and reduces the medical cost.

At present, a general medicine recommendation mode is usually to collect massive historical medical data and directly construct a medicine recommendation model by using the historical medical data. However, various types of noise inevitably exist in the medical data, such as the fact that a doctor opens the medical data by mistake, takes a medicine instead, and the like, so that the medicines in the medical data are not completely matched with the actual injury condition, and the accuracy of the constructed medicine recommendation model is low.

Disclosure of Invention

In view of the above, the application provides a general medication recommendation method, device, equipment and medium based on artificial intelligence, which can be used for solving the technical problem that the accuracy of a constructed medication recommendation model is low due to the fact that a general medication recommendation model is directly constructed by noisy historical medical data in the current general medication recommendation mode.

According to one aspect of the application, a general medication recommendation method based on artificial intelligence is provided, and the method comprises the following steps:

determining a sample set constructed from sample pathology data, the sample set comprising a first sample set in which no noise data exists after noise detection and a second sample set in which no noise detection exists;

pre-training a first medicine recommendation model by utilizing a first sample in the first sample set, and inputting a second sample in the second sample set into the pre-trained first medicine recommendation model to obtain a first medicine recommendation result;

screening a first preset number of second samples in the second sample set for noise detection according to the prediction probability and prediction uncertainty of the first medicine recommendation result, and iteratively training the first medicine recommendation model by using the second samples without noise data after noise detection and the first sample set to obtain a target medicine recommendation model;

acquiring patient pathological data uploaded by a target visiting patient, and inputting the patient pathological data into the target medication recommendation model to obtain a target medication recommendation result;

screening out a second preset number of target recommended medications meeting preset safe medication detection rules according to the target medication recommendation result, and sending the target recommended medications to the target visiting patients.

According to another aspect of the present application, there is provided an artificial intelligence based general medication recommendation device, comprising:

a determining module, configured to determine a sample set constructed from sample pathology data, including a first sample set in which no noise data exists after noise detection and a second sample set in which no noise detection exists;

the first training module is used for pre-training a first medicine recommendation model by utilizing a first sample in the first sample set, inputting a second sample in the second sample set into the pre-trained first medicine recommendation model, and acquiring a first medicine recommendation result;

the second training module is used for screening a first preset number of second samples in the second sample set for noise detection according to the prediction probability and prediction uncertainty of the first medicine recommendation result, and iteratively training the first medicine recommendation model by using the second samples without noise data after noise detection and the first sample set to obtain a target medicine recommendation model;

the input module is used for acquiring pathological data of a patient uploaded by a target visiting patient and inputting the pathological data of the patient into the target medication recommendation model to obtain a target medication recommendation result;

the first sending module is used for screening out a second preset number of target recommended medications meeting preset safe medication detection rules according to the target medication recommendation result and sending the target recommended medications to the target visiting patient.

According to yet another aspect of the present application, there is provided a storage medium having stored thereon a computer program which, when executed by a processor, implements the artificial intelligence based general medication recommendation method described above.

According to yet another aspect of the present application, there is provided a computer device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, the processor implementing the artificial intelligence based general medication recommendation method when executing the program.

By means of the technical scheme, compared with the current general medication recommending mode, the general medication recommending method, the device, the equipment and the medium based on the artificial intelligence can firstly determine a sample set constructed by sample pathological data, wherein the sample set comprises a first sample set without noise data after noise detection and a second sample set without noise detection; pre-training a first medicine recommendation model by using a first sample in a first sample set, and inputting a second sample in a second sample set after judging that the pre-training of the first medicine recommendation model is finished to obtain a first medicine recommendation result; screening a first preset number of second samples in a second sample set for noise detection according to the prediction probability and prediction uncertainty of the first medicine recommendation result, iteratively training a first medicine recommendation model by using the second samples without noise data after the noise detection and the first sample set until the training process meets a preset iteration termination condition, and judging that the iterative training of the first medicine recommendation model is finished to obtain a target medicine recommendation model; acquiring patient pathological data uploaded by a target visiting patient, and inputting the patient pathological data into a target medication recommendation model to obtain a target medication recommendation result; and screening out a second preset number of target recommended medications meeting the preset safe medication detection rule according to the target medication recommendation result, and sending the target recommended medications to the target visiting patients. Through the technical scheme in the application, the method and the device can be applied to an artificial intelligence technology, and on the basis of massive noise medical data, through iterative cleaning of noise data, the sample size is continuously enriched, and then repeated iterative training is performed on the medication recommendation model by using the updated sample set, so that a general medication recommendation model with high precision can be constructed. In addition, according to the method and the device, manual denoising processing is not required to be carried out on a large number of samples, the training cost of the medication recommendation model can be effectively saved, and the training efficiency is improved.

The foregoing description is only an overview of the technical solutions of the present application, and the present application can be implemented according to the content of the description in order to make the technical means of the present application more clearly understood, and the following detailed description of the present application is given in order to make the above and other objects, features, and advantages of the present application more clearly understandable.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application to the disclosed embodiment. In the drawings:

FIG. 1 is a flowchart illustrating a general medication recommendation method based on artificial intelligence according to an embodiment of the present application;

FIG. 2 is a flow chart of another artificial intelligence-based general medication recommendation method provided by an embodiment of the present application;

FIG. 3 is a schematic structural diagram illustrating an artificial intelligence-based general medication recommendation device according to an embodiment of the present application;

fig. 4 shows a schematic structural diagram of another artificial intelligence-based general medication recommending device provided by the embodiment of the application.

Detailed Description

The embodiment of the application can realize accurate recommendation of general medicine based on the artificial intelligence technology. Among them, Artificial Intelligence (AI) is a theory, method, technique and application system that simulates, extends and expands human Intelligence using a digital computer or a machine controlled by a digital computer, senses the environment, acquires knowledge and uses the knowledge to obtain the best result.

The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a robot technology, a biological recognition technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.

The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.

The technical problem of low accuracy of the existing general medicine recommending mode is solved. The application provides a general medication recommending method based on artificial intelligence, which comprises the following steps of:

101. a set of samples constructed from the sample pathology data is determined, the set of samples including a first set of samples that have been noise detected without noise data and a second set of samples that have not been noise detected.

Wherein each sample pathological data is marked with pathological information and a medication label determined according to the pathological information, and a sample set D_uSample pathology data 1: [ X ]₁,Y₁]Sample pathological data 2: [ X:₂,Y₂]sample pathological data 3: [ X:₃,Y₃]… sample pathological data N: [ X_N,Y_N]E.g. sample set D_uSample pathology data 1: [ common cold, Tanfei (oseltamivir phosphate capsule), ribavirin, amoxicillin, Yinqiao powder, lotus antipyretic capsule … Ankahuangmin capsule]Sample pathological data 2: [ acne, davidian (adapalene gel), isotretinoin, licorzinc particles … fusidic acid cream]The pathological data of the sample 3 are [ hepatitis B, entecavir, tenofovir and lamivudine … compound glycyrrhizin capsule]… pathological data of sample N: [ gastric ulcer, omeprazole, lansoprazole, pantoprazole, rabeprazole … esomeprazole: [ gastric ulcer, pantoprazole, gastric ulcer, pantoprazole, gastric ulcer, pantoprazole, rabeprazole … esomeprazole, gastric ulcer, and gastric ulcer]}. The first sample set is composed of first samples in which no noise data exists or noise data is removed, and the second sample set is composed of second samples in which it is uncertain whether noise data exists.

Due to the inevitable existence of various types of noise in the medical data, such as misopening by a doctor, taking a medicine instead, and the like, the medicine in the data is not completely matched with the actual injury condition, which affects the accuracy of the constructed recommendation model. Therefore, for this embodiment, a part of the first samples may be screened from the sample pathological data in advance, noise detection may be performed on the first samples, and when it is detected that no noise data exists in the first samples, the first samples may be stored in the first sample set; when the first sample is detected to have noise data, denoising processing is carried out on the first sample, and the first sample is also stored in the first sample set after denoising processing. Accordingly, second samples other than the first sample in the sample pathology data may be collectively stored in the second sample set. Through the method, the obtained first sample set has no noise data, so that the method can be applied to pre-training of the medication recommendation model to ensure the accuracy of the pre-training medication recommendation model. The second sample set with potential noise data can be used for carrying out iterative training on the pre-trained medication recommendation model, automatic cleaning of the noise data is achieved in the iterative training process, and model accuracy of the medication recommendation model can be guaranteed based on a large number of training samples while the training samples are enriched.

When the first sample is subjected to noise detection and denoising processing, as an optional mode, a preset number of sample pathological data can be extracted in a sample set, first pathological information in the sample pathological data and a first medication label determined according to the first pathological information are extracted, whether the first pathological information and the first medication label are matched with a preset pathological medication information table or not is judged through noise detection, if the first pathological information and the first medication label are matched with the preset pathological medication information table, it is determined that the sample pathological data do not need denoising processing, the sample pathological information is further stored in the first sample set, and if the first pathological information and the first medication label are not matched with each other, the first medication label is updated to a second medication label with the highest matching degree with the first pathological information according to the preset injury medication information table and stored in the first sample set.

As another optional mode, a preset number of sample pathological data can be extracted from the sample set, the sample pathological data is delivered to a preset management terminal for auditing and checking, the sample pathological data with wrong information is corrected, and the sample pathological data without information errors and the corrected sample pathological data are uniformly stored in a first sample as the first sampleThis is the focus. For example, given sample set D_uSample pathology data 1: [ X ]₁,Y₁]Sample pathological data 2: [ X:₂,Y₂]sample pathological data 3: [ X:₃,Y₃]… sample pathological data N: [ X_N,Y_N]Randomly sampling p% of sample pathological data from the sample pathological data, sending the sample pathological data to a preset management terminal, manually auditing by a specialist corresponding to the preset management terminal, checking whether a first medicine label of the p% of sample pathological data accords with first pathological information or not, obtaining after checking that q% of the p% of sample pathological data is correct and p% -q% of sample pathological data is incorrect, manually modifying the p% -q% of sample pathological data, and adding the q% of sample pathological data and the modified p% -q% of sample pathological data into a first sample set D₁1-p% of the sample pathology data without noise detection is stored in a second sample set D₂Sample set D_uFirst set of samples D₁Second set of samples D₂}。

Through the steps in the embodiment, part of sample pathological data in a sample set is extracted, noise detection is carried out on the sample pathological data, denoising processing is carried out on the sample pathological data with noise data, the existing noise data are eliminated, medicines in the sample pathological data are completely matched with the actual injury condition, a clean and noiseless training sample is provided for pre-training of the medication recommendation model, and interference of the noise data in the pre-training process of the medication recommendation model is avoided.

For the subject of execution of the present application, which may be a general medication recommendation device, configured on the client side or the server side, a sample set constructed from sample pathology data may be first determined, which includes a first sample set without noise data after noise detection and a second sample set without noise detection; the first sample in the first sample set is used for pre-training the first medicine recommendation model, and after the pre-training of the first medicine recommendation model is judged to be completed, the second sample in the second sample set is input to obtain a first medicine recommendation result; then screening a first preset number of second samples in a second sample set for noise detection according to the prediction probability and prediction uncertainty of the first medicine recommendation result, iteratively training a first medicine recommendation model by using the second samples without noise data after the noise detection and the first sample set until the training process meets a preset iteration termination condition, and judging that the iterative training of the first medicine recommendation model is finished to obtain a target medicine recommendation model; furthermore, the pathological data of the patients uploaded by the target visiting patients can be input into the target medication recommendation model to obtain target medication recommendation results, a second preset number of target recommended medications meeting the preset safe medication detection rules are screened out according to the target medication recommendation results, and the target recommended medications are sent to the target visiting patients.

102. And pre-training a first medicine recommendation model by using a first sample in the first sample set, and inputting a second sample in the second sample set into the pre-trained first medicine recommendation model to obtain a first medicine recommendation result.

The first medicine recommendation model may correspond to any one of existing Neural network models, for example, a Bayesian Neural network model (BNN), a linear regression model, a decision tree model, a Neural network model, a support vector machine model, a hidden markov model, and the like. In this embodiment, the first drug recommendation model may be a Bayesian Neural network model (BNN), which combines probabilistic modeling and a Neural network to predict not only the outcome, but also the prediction probability and prediction uncertainty of the prediction outcome, where the weight in the Bayesian Neural network model is a random variable W rather than a definite value, which is fundamentally different from a general Neural network, and the observed data, i.e., the training data D ═ X, Y, where X is the input data and Y is the label data, and the prediction outcome is represented by a distribution.

Because the prediction results are represented by distributions, the prediction probability of the prediction results and the prediction uncertainty can be obtained. The higher the prediction probability is, the higher the probability that the predicted medicine corresponds to the medicine recommended by the final model is, and the lower the prediction uncertainty is, the higher the prediction accuracy of the prediction probability is.

For this embodiment, the pre-training the first medication recommendation model by using the first sample in the first sample set may specifically include: the method comprises the steps of inputting first pathological information in a first sample as input features into a first medicine recommendation model, pre-training the first medicine recommendation model, judging that the pre-training of the first medicine recommendation model is finished when the similarity between a medicine recommendation result output by the first medicine recommendation model and a first medicine label corresponding to the first pathological information of the first sample is larger than a preset threshold value or a corresponding loss function is smaller than the preset threshold value, inputting a second sample in a second sample set into the pre-trained first medicine recommendation model, and outputting the prediction probability and prediction uncertainty of the first medicine recommendation result and the first medicine recommendation result.

103. And screening a first preset number of second samples in a second sample set for noise detection according to the prediction probability and prediction uncertainty of the first medicine recommendation result, and iteratively training a first medicine recommendation model by using the second samples without noise data after the noise detection and the first sample set to obtain a target medicine recommendation model.

For the embodiment, since the number of the first samples with the medication label is small, the pre-trained first medication recommendation model still cannot ensure high accuracy. In this regard, after the first medication recommendation result is obtained, a first preset number of second samples, which have corresponding prediction probabilities smaller than a first preset threshold and prediction inaccuracies larger than a second preset threshold, may be extracted from the second sample set according to the prediction probability and prediction uncertainty of the first medication recommendation result. Since the prediction probability corresponding to the extracted second sample is small and the prediction inaccuracy is large, the probability that the second sample has the noise data can be determined to be large, so that the noise detection can be performed on the second sample, the noise reduction processing can be performed on the second sample having the noise data, and after the noise reduction processing is performed, the second sample is configured with the correct second medication label. And then, a second sample configured with a second medication label is used for training the first medication recommendation model in an iterative mode until the first medication recommendation model meets a preset iteration termination condition, and the first medication recommendation model is determined to be a target medication recommendation model.

In particular, it can be done with pre-trainingAnd the first medicine recommending model identifies second samples in the second sample set which are not subjected to noise detection to obtain a first medicine recommending result. And then, successively screening second samples obviously having noisy data in a second sample set for denoising according to the prediction probability and prediction uncertainty of the first medicine recommendation result, storing the second samples which do not have noise and are configured with second medicine labels after denoising in the first sample set, and continuing to train the first medicine recommendation model pre-trained in the step 102 of the embodiment by using the updated first sample set. The method comprises the steps of inputting pathological information of a first sample or a second sample in a first sample set into a first medicine recommendation model as input features, performing enhancement training on the first medicine recommendation model, and judging that the enhancement training of the first medicine recommendation model is completed under the first sample set after current updating when the similarity between a medicine recommendation result output by the first medicine recommendation model and a first medicine label or a second medicine label is larger than a preset threshold value or a corresponding loss function is smaller than the preset threshold value. Further, the second sample of the screening part can be repeatedly executed to perform denoising processing, the denoised second sample is updated to the first sample set, the process of training the first medicine recommendation model by using the updated first sample set is used, the enhancement training of the first medicine recommendation model is continuously performed until the training process is judged to meet the preset iteration termination condition, and the first medicine recommendation model is judged to be finished through iteration training, so that the target medicine recommendation model is obtained. Wherein the preset iteration termination condition comprises: training a first medicine recommendation model to reach the maximum iteration number N; and/or the second sample set D obtained in the current iteration₂The prediction uncertainty (uncertaintiy) of all samples in (a) is less than a preset threshold.

In the above-mentioned screening of the first preset number of second samples from the second sample set for noise detection, and denoising in the presence of noise, the specific implementation process may refer to the description related to the noise detection and denoising for the first sample in the step 101 in the embodiment, and details are not repeated here.

For this embodiment, the second sample set D may be predicted using the first drug recommendation₂The second sample of (1), evaluating the prediction result with precision @ kDifference to the original tag result, precision @ k<The second sample of 0.5 is placed as a potentially noisy sample in the third sample set D₃. Further, a third sample set D can be obtained by using the first medicine recommendation model₃The uncertainty (uncertainties) of each second sample in the first sample set D is obtained, the q% second sample with the highest uncertainty is subjected to denoising processing by using the denoising processing method, for example, the denoising processing method can be audited and corrected by experts, and the denoised second sample is added into the first sample set D₁Simultaneously in a third sample set D₃Wherein the second sample is removed. Wherein precision @ k: for example, if the original label of a sample contains 5 drugs, k is 5, then precision @5 means that the 5 drugs with the highest score in the prediction result are taken out, and some of the 5 drugs are the same as the 5 drugs in the original label, so as to calculate the result, for example, if there are 3 pairs in the prediction result, precision @5 is 3/5 is 0.6.

Through the general medication recommending method based on artificial intelligence in the embodiment, the problem that the first sample volume in the first sample set is small and the first medication recommending model obtained through training according to the first sample is not accurate enough can be solved, noise detection is performed by screening a first preset number of second samples in the second sample set, noise removal processing is performed on the second samples with noise, the second samples without noise are stored in the first sample set, the sample volume in the first sample set is continuously increased, iterative training is performed on the first medication recommending model, and the target medication recommending model with high accuracy is obtained.

104. And acquiring the pathological data of the patient uploaded by the target visiting patient, and inputting the pathological data of the patient into the target medication recommendation model to obtain a target medication recommendation result.

The target medication recommendation model is the first medication recommendation model after the iterative training process in the embodiment step 103 meets the preset iteration termination condition. Continuously verifying a preset iteration termination condition in an iteration training process of the first medicine recommendation model, and when the first medicine recommendation model is judged to be trained to reach the iteration times N; and/or the prediction uncertainty (uncertainties) of all samples in the second sample set obtained in the current iteration is smaller than a preset threshold, the first medicine recommendation model in the current training stage can be determined to be a target medicine recommendation model which can be applied to an actual application scene and generates medicine recommendation according to pathological data of a patient. The target medication recommendation model may correspond to any one of the existing Neural network models, for example, a Bayesian Neural network model (BNN), a linear regression model, a decision tree model, a Neural network model, a support vector machine model, a hidden markov model, and the like. In this embodiment, the target medication recommendation model may be a Bayesian Neural network model (BNN) that is iteratively trained; the target visiting patient is a visiting patient in a general scene, and the pathological data of the patient can specifically comprise information such as diagnosis data and inspection and examination results.

For the embodiment, as a preferred mode, when uploading the pathological data of the patient, the pathological data of the patient includes a mandatory item and a optional item, after receiving the pathological data of the patient uploaded by the target visiting patient, the pathological data of the patient can be firstly subjected to deficiency inspection of the mandatory item data, if the pathological data of the patient is judged to have deficiency of the mandatory item data, the pathological data of the patient can be subjected to data filling processing according to a preset data filling rule, or the reason that the target medication recommendation result is not obtained is sent to the target visiting patient, and the target visiting patient is prompted to supplement and upload the deficient data; after the patient pathological data is determined to be complete, the following embodiment steps of generating medication recommendation results according to the patient pathological data are further executed. In a specific application scenario, when the reason that the target medication recommendation result is not obtained is sent to the target visiting patient and the target visiting patient is prompted to supplement and upload the missing data, the index instruction corresponding to the missing data can be output, so that the target visiting patient can effectively fill the missing data according to the index instruction. The index instructions can include data standard examples corresponding to the required items and the optional items and index analysis, so that the uploaded pathological data of the patient are more comprehensive, and the obtained target medication recommendation result is more accurate. For example: mandatory item 1: age (e.g.: 10; index: check contraindications for taking medicines at different ages), mandatory item 2: allergy medication (e.g. "penicillin"; index: look at drugs that may cause allergy), mandatory item 3: whether or not to be pregnant (such as yes or no; index: check pregnancy contraindication), etc.; and (4) selecting the filling item 1: name (e.g. "Zhang three"; index: for verifying patient identity), option 2: gender (e.g. "male" or "female"; index: looking at diseases that may arise from different genders), etc.

105. And screening out a second preset number of target recommended medications meeting the preset safe medication detection rule according to the target medication recommendation result, and sending the target recommended medications to the target visiting patients.

For the embodiment, on the basis that the target medication recommendation result is obtained by the target medication recommendation model, the screening is performed according to the preset safe medication detection rule again, including screening out forbidden drugs proposed by the national drug administration, and providing the optimal medication combination, which may include lowest-cost combined drugs, fastest-effect combined drugs, minimum-irritation combined drugs, and the like, and sending the optimal medication combination to the target visiting patient for the target visiting patient to select according to the self-demand. If the target visiting patient uploads influenza plus fever of 39 degrees, is 22 years old, has no necessary information such as pregnancy, the target medication recommendation result gives 10 medicines, 8 medicines meeting the rules are screened out according to the preset safety medication detection rules, such as screening analgin injection, analgin chlorpromazine injection, analgin drop, analgin nasal drop, analgin solution tablet for nasal drop, vitamin C Yinqiao tablet and the like which are forbidden by the national medicine supervision bureau, and the lowest cost combination is given out: the compound paracetamol and amantadine hydrochloride tablet and the isatis root granules have the fastest combination effect: ibuprofen suspension and copperleaf antipyretic, minimal irritative combination: acetaminophen (paracetamol) and vitamin C.

Through the general medicine recommendation method based on artificial intelligence in the embodiment, different requirements of different patients under the same or similar pathological information are met, and the precision of general medicine recommendation is improved.

By the general medication recommendation method based on artificial intelligence in the embodiment, a sample set constructed by sample pathological data can be determined firstly, wherein the sample set comprises a first sample set without noise data after noise detection and a second sample set without noise detection; pre-training a first medicine recommendation model by using a first sample in a first sample set, and inputting a second sample in a second sample set after judging that the pre-training of the first medicine recommendation model is finished to obtain a first medicine recommendation result; screening a first preset number of second samples in a second sample set for noise detection according to the prediction probability and prediction uncertainty of the first medicine recommendation result, iteratively training a first medicine recommendation model by using the second samples without noise data after the noise detection and the first sample set until the training process meets a preset iteration termination condition, and judging that the iterative training of the first medicine recommendation model is finished to obtain a target medicine recommendation model; acquiring patient pathological data uploaded by a target visiting patient, and inputting the patient pathological data into a target medication recommendation model to obtain a target medication recommendation result; and screening out a second preset number of target recommended medications meeting the preset safe medication detection rule according to the target medication recommendation result, and sending the target recommended medications to the target visiting patients. The method can be applied to an artificial intelligence technology, and on the basis of massive noise medical data, through iterative cleaning of the noise data, the sample size is continuously enriched, and then repeated iterative training is performed on the medication recommendation model by using the updated sample set, so that a general medication recommendation model with high accuracy can be constructed. In addition, according to the method and the device, manual denoising processing is not required to be carried out on a large number of samples, the training cost of the medication recommendation model can be effectively saved, and the training efficiency is improved.

Further, as a refinement and an extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process in this embodiment, another general medication recommendation method based on artificial intelligence is provided, as shown in fig. 2, and the method includes:

201. determining a sample set constructed by sample pathological data, wherein the sample set comprises a first sample set subjected to denoising processing and a second sample set which is not subjected to denoising processing.

The specific implementation process may refer to the related description in step 101 of the embodiment, and is not described herein again.

202. And pre-training a first medicine recommendation model by using a first sample in the first sample set, and inputting a second sample in the second sample set into the pre-trained first medicine recommendation model to obtain a first medicine recommendation result.

The specific implementation process may refer to the related description in step 102 of the embodiment, and is not described herein again.

203. And according to the first medicine recommending result, extracting a first preset number of second samples with the corresponding prediction probability smaller than a first preset threshold and the prediction inaccuracy larger than a second preset threshold from a second sample set.

For this embodiment, after the first medicine recommendation result corresponding to the second sample is obtained, a first preset number of second samples, of which the corresponding prediction probabilities are smaller than a first preset threshold and the prediction inaccuracy is larger than a second preset threshold, may be extracted from the second sample set further according to the prediction probability and the prediction uncertainty of the first medicine recommendation result. Since the prediction probability of the extracted second sample is small and the prediction inaccuracy is large, it can be determined that the second sample has a high probability of having noise data, and the noise influence in the second sample set can be eliminated by performing noise detection and noise processing on the second sample. The first preset threshold and the second preset threshold are both values between 0 and 1, and specific values can be set according to actual application scenarios.

For example, if there are 100 second samples in the second sample set, all the second samples are input into the pre-trained first drug recommendation model, and the first drug recommendation results corresponding to the 100 second samples, the prediction probability P and the prediction uncertainty 1- α of the first drug recommendation results are obtained according to the first drug recommendation model, where the preset screening condition is: and (3) screening 100 second samples to obtain 30 samples meeting the conditions, wherein the prediction probability P < a first preset threshold value 0.8 and a second preset threshold value > and the prediction inaccuracy (1-alpha) is 0.2.

204. Extracting pathological information and a first medicine label corresponding to the second sample, judging whether the pathological information is matched with the first medicine label or not according to a preset injury medicine information table, if the pathological information is not matched with the first medicine label, updating the first medicine label into a second medicine label with the highest matching degree with the pathological information according to the preset injury medicine information table, acquiring the second sample without noise data after noise detection, updating the second sample into a first sample set, and training the first medicine recommendation model by utilizing the updated first sample set.

For this embodiment, specifically, noise detection and noise processing can be performed on the second sample extracted through the screening condition, that is, whether pathological information is matched with the first medical label or not is judged according to the preset injury medication information table, and expert review or artificial intelligence review can be adopted in the process, and if the pathological information is not matched with the first medical label, the existence of noise is determined. Furthermore, denoising processing can be performed on a second sample with noise, the first medicine label is updated to a second medicine label with the highest matching degree with pathological information according to a preset injury medicine information table, the first medicine label is removed, denoising processing is confirmed to be completed, the second sample without noise after noise detection is updated to a first sample set, and the updated first sample set is used for training a first medicine recommendation model. For a specific implementation process, reference may be made to the related description of performing noise detection and denoising processing on the first sample in step 101 in the embodiment, and details are not described here again.

For example, based on the example in the embodiment step 203, after 30 second samples meeting the preset screening condition are subjected to noise detection, 10 second samples containing noise are subjected to denoising processing, the denoised 10 second samples containing no noise and the remaining 20 second samples are updated to the first sample set to serve as data for iterative training of the first drug recommendation model, and the updated first sample set is used for training the first drug recommendation model.

By the method, the amount of the noisy sample data of the first medicine recommendation model for iterative training is increased, and the target recommendation model obtained through training is more accurate.

205. And repeatedly executing the process of screening the first preset number of second samples for denoising, updating the denoised second samples to the first sample set, training the first medicine recommendation model by using the updated first sample set until the training process is determined to meet the preset iteration termination condition, and judging that the iterative training of the first medicine recommendation model is finished to obtain the target medicine recommendation model.

For the embodiment, when the training process is judged to meet the preset iteration termination condition, the iterative training of the first medicine recommendation model can be ended, and the first medicine recommendation model in the current training process is determined to be the application model which can be finally applied to the actual general medicine recommendation scene.

206. And acquiring the pathological data of the patient uploaded by the target visiting patient, and inputting the pathological data of the patient into the target medication recommendation model to obtain a target medication recommendation result.

The target visiting patient is a visiting patient in a general scene, and the pathological data of the patient specifically comprises information such as diagnosis data and examination results.

In a specific application scenario, before the step of this embodiment is executed, as an optional implementation, the method further includes: determining the disease level corresponding to a target visiting patient according to the pathological data of the patient, and judging whether the disease level triggers a drug-assisted treatment condition; correspondingly, if the condition level is judged to trigger the drug-assisted treatment condition, the pathological data of the patient is input into the target medication recommendation model to obtain a target medication recommendation result. The cases where it is determined that the condition for drug-assisted therapy is not triggered include: the disease can be cured without drug treatment, if the body temperature does not exceed 37.5 degrees, the patient does not take the drug as much as possible to defervesce if the fever occurs; the disease condition is too serious, the treatment can not be carried out by only using the medicine, the operation treatment is needed, for example, sudden death, respiratory and heartbeat arrest and the like are taken as the manifestations when the pulmonary embolism occurs, and the first-aid treatment is needed at the moment; or the medicines prescribed for the symptoms are prescription medicines and the like, and the medicines are prescribed by doctors, such as powerful hypnotics, tranquilizers and the like.

Correspondingly, before the step of this embodiment is executed, as another optional implementation, the method further includes: judging whether the pathological data of the patient has data loss or not through data analysis of the pathological data of the patient; correspondingly, if the pathological data of the patient is judged to have data missing, data filling processing is carried out on the injury data of the patient according to a preset data filling rule, or a target visiting patient is prompted to supplement and upload the missing injury data of the patient, and the injury data of the patient is determined to be complete.

207. And screening out a second preset number of target recommended medications meeting the preset safe medication detection rule according to the target medication recommendation result, and sending the target recommended medications to the target visiting patients.

As to this embodiment, as an optional implementation manner, step 207 of this embodiment may specifically include: performing medicine safety detection on combined medicines and/or single medicines with recommended scores larger than a preset threshold value in the target medicine recommendation results according to a plurality of preset safety medicine detection rules to obtain a target recommended medicine list passing the medicine safety detection, wherein the combined medicines are a combination at least comprising two single medicines; and sending the target recommended medication list to the target visiting patient.

For this embodiment, when performing drug security detection on a combined drug and/or a single drug whose recommended score is greater than a preset threshold in a target medication recommendation result according to a plurality of preset security medication detection rules and obtaining a target recommended medication list passing the drug security detection, the steps of the embodiment may specifically include: carrying out drug interaction detection on the combined drug with the recommended value larger than a third preset threshold value according to a preset drug interaction detection rule to obtain a first detection result; performing contraindication detection on the combined medicine and the single medicine with the recommended values larger than a third preset threshold value according to a preset contraindication detection rule to obtain a second detection result; carrying out drug allergy detection on the combined drug and the single drug with the recommended values larger than a third preset threshold value according to a preset drug allergy detection rule to obtain a third detection result; and determining a target recommended medication list suitable for the target visiting patient according to the first detection result, the second detection result and the third detection result.

In a specific application scenario, after sending the target recommended medication to the target visiting patient, as an optional way to help the target visiting patient to scientifically select and take the medication, the method may further include: receiving a selection instruction of a target visiting patient for target recommended medication; and responding to the selection instruction, acquiring the medicine information of the selected target recommended medication of the target visiting patient in the medicine knowledge graph, and sending the medicine information to the target visiting patient, wherein the medicine information can comprise medicine selling price information, medicine component information, medicine taking contraindication information, medicine taking dosage information and the like.

Accordingly, as another alternative, after obtaining the target recommended medication list, the method may further include: the method comprises the steps of obtaining names of a plurality of selling points of each medicine in a target recommended medication list and position coordinates of each selling point, obtaining position coordinates of a starting point of a target patient, calculating the distance between the position coordinates of the starting point of the target patient and each selling point of each medicine, and displaying the names of the plurality of selling points of each medicine in the target recommended medication list and the distance between a target visiting patient and each selling point of the medicine.

Further, as a specific implementation of the method shown in fig. 1, an embodiment of the present application provides a general medication recommendation device based on artificial intelligence, as shown in fig. 3, the device includes: a determining module 31, a first training module 32, a second training module 33, an input module 34, and a first transmitting module 35;

a determining module 31, operable to determine a sample set constructed from the sample pathology data, including a first sample set in which no noise data exists after noise detection and a second sample set in which no noise data exists;

the first training module 32 may be configured to pre-train a first medication recommendation model using a first sample in the first sample set, and input a second sample in the second sample set into the pre-trained first medication recommendation model to obtain a first medication recommendation result;

the second training module 33 is configured to screen a first preset number of second samples in the second sample set according to the prediction probability and the prediction uncertainty of the first medication recommendation result to perform noise detection, and iteratively train the first medication recommendation model by using the second samples without noise data after the noise detection and the first sample set to obtain a target medication recommendation model;

the input module 34 is used for acquiring pathological data of a patient uploaded by a target visiting patient, and inputting the pathological data of the patient into a target medication recommendation model to obtain a target medication recommendation result;

the first sending module 35 may be configured to screen out a second preset number of target recommended medications meeting the preset safe medication detection rule according to the target medication recommendation result, and send the target recommended medications to the target visiting patient.

In a specific application scenario, in order to screen a first preset number of second samples from the second sample set for noise detection according to the prediction probability and prediction uncertainty of the first medicine recommendation result, the second training module 33 is specifically configured to extract a first preset number of second samples, of which the corresponding prediction probability is smaller than a first preset threshold and the prediction inaccuracy is greater than a second preset threshold, from the second sample set according to the first medicine recommendation result; extracting pathological information and a first medicine label corresponding to the second sample, and judging whether the pathological information is matched with the first medicine label or not according to a preset injury medicine information table; and if the pathological information is not matched with the first medicine label, updating the first medicine label into a second medicine label with the highest matching degree with the pathological information according to a preset injury medicine information table.

Correspondingly, when the first medication recommendation model is iteratively trained by using the first sample set and the second sample without noise data after noise detection to obtain the target medication recommendation model, the second training module 33 may be specifically configured to obtain the second sample without noise data after noise detection, update the second sample to the first sample set, and train the first medication recommendation model by using the updated first sample set; and repeatedly executing the process of screening the first preset number of second samples for noise detection, updating the second samples without noise data after the noise detection to the first sample set, training the first medicine recommendation model by using the updated first sample set until the training process is determined to meet the preset iteration termination condition, and judging that the iterative training of the first medicine recommendation model is finished to obtain the target medicine recommendation model.

In a specific application scenario, in order to determine the drug-assisted treatment condition of the patient pathology data before inputting the patient pathology data into the target medication recommendation model, as shown in fig. 4, the apparatus further includes: a decision module 36;

a determination module 36, configured to determine a disease level corresponding to the target visiting patient according to the pathological data of the patient, and determine whether the disease level triggers a drug-assisted therapy condition;

the input module 34 may be configured to input pathological data of the patient into the target medication recommendation model to obtain a target medication recommendation result if it is determined that the disease condition level triggers the medication auxiliary treatment condition.

In a specific application scenario, when a second preset number of target recommended medications meeting preset safe medication detection rules are screened out according to the target medication recommendation result and the target recommended medications are sent to target visiting patients, the first sending module 35 is specifically configured to perform drug safety detection on combined drugs and/or single drugs with recommendation scores larger than a preset threshold value in the target medication recommendation result according to a plurality of preset safe medication detection rules, and obtain a target recommended medication list passing the drug safety detection, wherein the combined drugs are a combination at least including two single drugs; and sending the target recommended medication list to the target visiting patient.

Correspondingly, in order to obtain a target recommended medication list passing through the drug safety detection, the first sending module 35 is specifically configured to perform drug interaction detection on a combined drug of which the recommended score is greater than a third preset threshold according to a preset drug interaction detection rule, so as to obtain a first detection result; performing contraindication detection on the combined medicine and the single medicine with the recommended values larger than a third preset threshold value according to a preset contraindication detection rule to obtain a second detection result; carrying out drug allergy detection on the combined drug and the single drug with the recommended values larger than a third preset threshold value according to a preset drug allergy detection rule to obtain a third detection result; and determining a target recommended medication list suitable for the target visiting patient according to the first detection result, the second detection result and the third detection result.

In a specific application scenario, in order to output medicine information of a selected medicine based on a selection instruction of a target visiting patient for target recommended medication after generating the target recommended medication, and further ensure medication safety of the patient, as shown in fig. 4, the apparatus further includes: a receiving module 37, a second transmitting module 38;

the receiving module 37 is configured to receive a selection instruction of the target visiting patient for the target recommended medication;

and the second sending module 38 is configured to, in response to the selection instruction, obtain, in the drug knowledge graph, drug information of the selected target recommended medication of the target visiting patient, and send the drug information to the target visiting patient.

It should be noted that other corresponding descriptions of the functional units related to the general medication recommending apparatus based on artificial intelligence provided in this embodiment may refer to the corresponding descriptions in fig. 1 to fig. 2, and are not described herein again.

Based on the methods shown in fig. 1 to fig. 2, correspondingly, the present embodiment further provides a storage medium, which may be volatile or nonvolatile, and on which computer readable instructions are stored, and when the computer readable instructions are executed by a processor, the method for recommending a general medication based on artificial intelligence shown in fig. 1 to fig. 2 is implemented.

Based on such understanding, the technical solution of the present application may be embodied in the form of a software product, which may be stored in a storage medium (which may be a CD-ROM, a usb disk, a removable hard disk, or the like), and includes several instructions to enable a computer device (which may be a personal computer, a server, or a network device, or the like) to execute the method of the embodiments of the present application.

Based on the method shown in fig. 1 to fig. 2 and the virtual device embodiments shown in fig. 3 and fig. 4, in order to achieve the above object, the present embodiment further provides a computer device, where the computer device includes a storage medium and a processor; a storage medium for storing a computer program; a processor for executing a computer program to implement the artificial intelligence based general medication recommendation method as described above with reference to fig. 1-2.

Optionally, the computer device may further include a user interface, a network interface, a camera, Radio Frequency (RF) circuitry, a sensor, audio circuitry, a WI-FI module, and so forth. The user interface may include a Display screen (Display), an input unit such as a keypad (Keyboard), etc., and the optional user interface may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface), etc.

It will be understood by those skilled in the art that the present embodiment provides a computer device structure that is not limited to the physical device, and may include more or less components, or some components in combination, or a different arrangement of components.

The storage medium may further include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the computer device described above, supporting the operation of information handling programs and other software and/or programs. The network communication module is used for realizing communication among components in the storage medium and communication with other hardware and software in the information processing entity device.

Through the above description of the embodiments, those skilled in the art will clearly understand that the present application can be implemented by software plus a necessary general hardware platform, and can also be implemented by hardware.

By applying the technical scheme of the application, compared with the prior art, the application can firstly determine a sample set constructed by sample pathological data, wherein the sample set comprises a first sample set without noise data after noise detection and a second sample set without noise detection; pre-training a first medicine recommendation model by using a first sample in a first sample set, and inputting a second sample in a second sample set after judging that the pre-training of the first medicine recommendation model is finished to obtain a first medicine recommendation result; screening a first preset number of second samples in a second sample set for noise detection according to the prediction probability and prediction uncertainty of the first medicine recommendation result, iteratively training a first medicine recommendation model by using the second samples without noise data after the noise detection and the first sample set until the training process meets a preset iteration termination condition, and judging that the iterative training of the first medicine recommendation model is finished to obtain a target medicine recommendation model; acquiring patient pathological data uploaded by a target visiting patient, and inputting the patient pathological data into a target medication recommendation model to obtain a target medication recommendation result; and screening out a second preset number of target recommended medications meeting the preset safe medication detection rule according to the target medication recommendation result, and sending the target recommended medications to the target visiting patients. Through the technical scheme in the application, the method and the device can be applied to an artificial intelligence technology, and on the basis of massive noise medical data, through iterative cleaning of noise data, the sample size is continuously enriched, and then repeated iterative training is performed on the medication recommendation model by using the updated sample set, so that a general medication recommendation model with high precision can be constructed. In addition, according to the method and the device, manual denoising processing is not required to be carried out on a large number of samples, the training cost of the medication recommendation model can be effectively saved, and the training efficiency is improved.

Those skilled in the art will appreciate that the figures are merely schematic representations of one preferred implementation scenario and that the blocks or flow diagrams in the figures are not necessarily required to practice the present application. Those skilled in the art will appreciate that the modules in the devices in the implementation scenario may be distributed in the devices in the implementation scenario according to the description of the implementation scenario, or may be located in one or more devices different from the present implementation scenario with corresponding changes. The modules of the implementation scenario may be combined into one module, or may be further split into a plurality of sub-modules.

The above application serial numbers are for description purposes only and do not represent the superiority or inferiority of the implementation scenarios. The above disclosure is only a few specific implementation scenarios of the present application, but the present application is not limited thereto, and any variations that can be made by those skilled in the art are intended to fall within the scope of the present application.

Claims

1. A general medication recommending method based on artificial intelligence is characterized by comprising the following steps:

2. The method of claim 1, wherein the screening a first predetermined number of second samples in the second sample set for noise detection based on the prediction probability and prediction uncertainty of the first medication recommendation comprises:

according to the first medicine recommending result, extracting a first preset number of second samples with corresponding prediction probabilities smaller than a first preset threshold and prediction inaccuracy larger than a second preset threshold from the second sample set;

extracting pathological information and a first medicine label corresponding to the second sample, and judging whether the pathological information is matched with the first medicine label or not according to a preset injury medicine information table;

and if the pathological information is not matched with the first medicine label, updating the first medicine label into a second medicine label with the highest matching degree with the pathological information according to the preset injury medicine information table.

3. The method of claim 2, wherein iteratively training the first medication recommendation model using the second sample and the first set of samples in the absence of noisy data after noise detection to obtain a medication recommendation model for the target comprises:

acquiring a second sample without noise data after noise detection, updating the second sample to the first sample set, and training the first medicine recommendation model by using the updated first sample set;

and repeatedly executing the process of screening the first preset number of second samples for noise detection, updating the second samples without noise data after the noise detection to the first sample set, training the first medicine recommendation model by using the updated first sample set until the training process is determined to meet the preset iteration termination condition, and judging that the iterative training of the first medicine recommendation model is finished to obtain the target medicine recommendation model.

4. The method of claim 1, prior to entering the patient pathology data into the medication target recommendation model to obtain a medication target recommendation, further comprising:

determining a disease level corresponding to the target visiting patient according to the pathological data of the patient, and judging whether the disease level triggers a drug-assisted treatment condition;

and if the condition level is judged to trigger the drug auxiliary treatment condition, inputting the pathological data of the patient into the target medication recommendation model to obtain a target medication recommendation result.

5. The method of claim 1, wherein the screening out a second preset number of target recommended medications meeting preset safe medication detection rules according to the target medication recommendation result and sending the target recommended medications to the target visiting patient comprises:

performing drug safety detection on combined drugs and/or single drugs with recommended scores larger than a preset threshold value in the target drug recommendation result according to a plurality of preset safety drug detection rules to obtain a target recommended drug list passing the drug safety detection, wherein the combined drugs are a combination at least comprising two single drugs;

and sending the target recommended medication list to the target visiting patient.

6. The method according to claim 5, wherein the step of performing drug safety detection on the combined drug and/or the single drug with the recommended score larger than a preset threshold in the target drug recommendation result according to a plurality of preset safety drug detection rules to obtain a target recommended drug list passing the drug safety detection comprises:

carrying out drug interaction detection on the combined drug with the recommended value larger than a third preset threshold value according to a preset drug interaction detection rule to obtain a first detection result;

performing contraindication detection on the combined medicine and the single medicine with the recommended values larger than a third preset threshold value according to a preset contraindication detection rule to obtain a second detection result;

carrying out drug allergy detection on the combined drug and the single drug with the recommended values larger than a third preset threshold value according to a preset drug allergy detection rule to obtain a third detection result;

determining a target recommended medication list applicable to the target visiting patient according to the first detection result, the second detection result and the third detection result.

7. The method of claim 1, further comprising:

receiving a selection instruction of the target visiting patient for the target recommended medication;

and responding to the selection instruction, acquiring the medicine information of the selected target recommended medication of the target visiting patient from a medicine knowledge graph, and sending the medicine information to the target visiting patient.

8. A medicine recommendation device is used in general branch of academic or vocational study based on artificial intelligence, its characterized in that includes:

9. A storage medium having stored thereon a computer program, which when executed by a processor implements the artificial intelligence based general medication recommendation method of any one of claims 1 to 7.

10. A computer device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor implements the artificial intelligence based general medication recommendation method of any one of claims 1-7 when executing the program.